Core concepts

Flireo AI enables you to build advanced phone agents that hold natural conversations, automate workflows, and integrate with your existing systems. To build effective agents, it's essential to understand the core components of our platform, the different agent types, and the orchestration model behind every call.

The Flireo Orchestration Model

Every phone call with a Flireo agent follows a structured pipeline designed for natural, responsive voice interaction:

Speech-to-Text (STT):
When a user speaks to your agent, our speech recognition engine converts the audio into accurate text.

Conversational AI (LLM):
The transcribed text is passed to a large language model (LLM). The LLM interprets user intent, manages conversation state, and determines the next action. This may include:

Generating a direct response.

Executing a function (e.g., sending an email, querying a database, booking an appointment).

Both responding and executing actions.

Text-to-Speech (TTS):
The agent’s reply—whether generated by the LLM or by a function—is passed to the text-to-speech engine, which converts it into lifelike audio. This is played back to the caller in real time.

Pipeline Summary:

Caller speaks → [STT] → Text → [LLM] → Response/Function → [TTS] → Spoken reply

Prompting Best Practices

Because Flireo agents communicate over the phone, what they say must sound natural when spoken aloud. When you design prompts and instructions for your agents:

Write full sentences and paragraphs.
Avoid bullet lists, note-style fragments, or text message formatting.

Use conversational language.
Pretend you’re scripting a radio host, not writing a chat bot.

Test responses aloud.
If it sounds awkward when spoken, rewrite it for clarity and flow.

Tip: If your agent's instructions use short, choppy sentences or bullet points, the TTS will read these out literally, making the conversation sound robotic or unnatural.

Agent Types

Flireo supports three distinct agent architectures:

1. Default Agents

Description:
Use only Flireo's built-in tools and integrations (like sending emails, browsing the web, etc.).

Management:
Fully configured and managed in the Flireo dashboard.
Visual function editor available for customizing behavior.

Use case:
Fast deployment for common workflows, internal automations.

2. External Agents

Description:
Agents whose orchestration and logic are managed by your own backend server.
Flireo communicates with your server via secure webhooks during calls.

Event Types:

assistant-request:
When a call starts, Flireo sends a request for agent instructions/configuration.

tool-calls:
When the agent needs to execute a function (e.g., fetch data, trigger an API), Flireo sends a request for your server to process and respond.

end-of-call-report:
At the end of the call, Flireo sends a detailed report (transcript, analysis, metadata) for logging, analytics, or CRM integration.

status-updates:
Real-time notifications for call lifecycle events (started, finished, errors).

Configuration:
Functions can be defined visually in the dashboard under /assistants/[id]/functions, but execution is handled by your backend.

Use case:
Deeply integrated workflows, dynamic personalization, custom business logic.

3. Custom Agents

Description:
Fully bespoke agents built and managed by the Flireo team, with custom integrations on the Flireo infrastructure.

Management:
Functions and integrations are handled by Flireo; dashboard editing is restricted.

Use case:
Enterprise clients requiring advanced features, regulatory compliance, or integrations not possible with default/external agents.

Phone Number Management & Requirement Groups

Phone Numbers:
Agents need dedicated phone numbers to make and receive calls.

Standard numbers are turnkey and assigned automatically.

Regional numbers (with a specific area code) require verification and may have additional regulatory requirements.

Requirement Groups:
For compliance and verification, regional numbers require a set of requirements to be fulfilled.
These are managed through requirement groups, which track status and progress via API and dashboard.

Building Effective Voice Agents

Understand your agent type and its capabilities.

Design prompts for spoken conversation, not text chat.

Use Flireo's orchestration model to integrate functions, external logic, and real-time data for a seamless caller experience.

Take advantage of analytics and reporting to refine your agents over time.

Next Steps:
Learn how to create your first agent, connect phone numbers, and integrate your business logic with Flireo AI.

The Flireo Orchestration Model#

Prompting Best Practices#

Agent Types#

1. Default Agents#

2. External Agents#

3. Custom Agents#

Phone Number Management & Requirement Groups#

Building Effective Voice Agents#